Morphological operations (erosion, dilation, opening, closing) in Computer Vision

Introduction

Morphological operations help clean and change shapes in images by adding or removing pixels. They make it easier to find important parts or fix small mistakes in pictures.

Removing small noise dots from a scanned document image.

Making objects in a photo thicker or thinner to see their shape better.

Separating two objects that are touching in a picture.

Filling small holes inside objects in a black and white image.

Smoothing the edges of shapes in a binary image for better analysis.

Syntax

Computer Vision

cv2.erode(image, kernel, iterations=1)
cv2.dilate(image, kernel, iterations=1)
cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel)
cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)

image: Input image, usually black and white (binary) or grayscale.

kernel: A small matrix (like 3x3) that defines the shape of the operation.

iterations: How many times to apply the operation.

Examples

This makes the white parts in the image smaller by removing pixels at the edges.

Computer Vision

kernel = np.ones((3,3), np.uint8)
eroded = cv2.erode(image, kernel, iterations=1)

This makes the white parts bigger by adding pixels around edges, done twice.

Computer Vision

kernel = np.ones((5,5), np.uint8)
dilated = cv2.dilate(image, kernel, iterations=2)

Opening removes small white noise by eroding then dilating.

Computer Vision

kernel = np.ones((3,3), np.uint8)
opened = cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel)

Closing fills small black holes by dilating then eroding.

Computer Vision

kernel = np.ones((3,3), np.uint8)
closed = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)

Sample Model

This code creates a small black and white image with a white square and one noise pixel. It then shows how erosion, dilation, opening, and closing change the image.

Computer Vision

import cv2
import numpy as np

# Create a simple binary image with noise
image = np.zeros((7,7), dtype=np.uint8)
image[2:5, 2:5] = 255  # a white square
image[1,1] = 255      # noise pixel

kernel = np.ones((3,3), np.uint8)

# Apply erosion
eroded = cv2.erode(image, kernel, iterations=1)

# Apply dilation
dilated = cv2.dilate(image, kernel, iterations=1)

# Apply opening (remove noise)
opened = cv2.morphologyEx(image, cv2.MORPH_OPEN, kernel)

# Apply closing (fill holes)
closed = cv2.morphologyEx(image, cv2.MORPH_CLOSE, kernel)

print("Original image:\n", image)
print("Eroded image:\n", eroded)
print("Dilated image:\n", dilated)
print("Opened image:\n", opened)
print("Closed image:\n", closed)

Important Notes

Erosion shrinks white areas, useful to remove small white noise.

Dilation grows white areas, useful to fill small black holes.

Opening is erosion followed by dilation, good for removing noise without changing shape much.

Closing is dilation followed by erosion, good for filling holes inside objects.

Summary

Morphological operations change image shapes by adding or removing pixels.

Erosion shrinks white parts; dilation grows them.

Opening removes noise; closing fills holes.

Practice

1. What does the erosion operation do to the white parts of a binary image?

easy

A. It grows the white parts by adding pixels at the edges.

B. It removes noise by smoothing the edges.

C. It fills small holes inside the white parts.

D. It shrinks the white parts by removing pixels at the edges.

Solution

1. Step 1: Understand erosion effect on white pixels

   Erosion removes pixels from the boundaries of white regions, making them smaller.

2. Step 2: Compare with other operations

   Dilation grows white parts, opening removes noise, closing fills holes, so erosion must shrink white parts.

3. Final Answer:

   It shrinks the white parts by removing pixels at the edges. -> Option D

4. Quick Check:

   Erosion = Shrinks white parts [OK]

Hint: Erosion shrinks white areas by cutting edges [OK]

Common Mistakes:

• Confusing erosion with dilation
• Thinking erosion fills holes
• Mixing erosion with noise removal

2. Which of the following is the correct syntax to perform dilation using OpenCV in Python on an image img with a 3x3 kernel?

easy

A. cv2.erode(img, np.ones((3,3), np.uint8))

B. cv2.dilate(img, np.ones((3,3), np.uint8))

C. cv2.dilate(img, (3,3))

D. cv2.dilate(img, kernel=3)

Solution

1. Step 1: Recall dilation syntax in OpenCV

   Dilation requires the image and a structuring element (kernel), usually created with np.ones of desired size and type.

2. Step 2: Check options for correct usage

   cv2.dilate(img, np.ones((3,3), np.uint8)) uses cv2.dilate with a 3x3 kernel created by np.ones and correct dtype, which is valid syntax.

3. Final Answer:

   cv2.dilate(img, np.ones((3,3), np.uint8)) -> Option B

4. Quick Check:

   Dilation syntax = cv2.dilate(image, kernel) [OK]

Hint: Use np.ones((3,3), np.uint8) as kernel for dilation [OK]

Common Mistakes:

• Using erode instead of dilate
• Passing kernel size tuple directly
• Using wrong kernel datatype

3. Given the following Python code using OpenCV:

import cv2
import numpy as np
img = np.array([[0,0,0,0,0],
                [0,255,255,255,0],
                [0,255,0,255,0],
                [0,255,255,255,0],
                [0,0,0,0,0]], dtype=np.uint8)
kernel = np.ones((3,3), np.uint8)
eroded = cv2.erode(img, kernel)
print(eroded)

What will be the printed output?

medium

A. [[ 0 0 0 0 0]
[ 0 255 0 255 0]
[ 0 0 0 0 0]
[ 0 255 0 255 0]
[ 0 0 0 0 0]]

B. [[ 0 0 0 0 0]
[ 0 255 255 255 0]
[ 0 255 255 255 0]
[ 0 255 255 255 0]
[ 0 0 0 0 0]]

C. [[ 0 0 0 0 0]
[ 0 0 0 0 0]
[ 0 0 0 0 0]
[ 0 0 0 0 0]
[ 0 0 0 0 0]]

D. [[255 255 255 255 255]
[255 255 255 255 255]
[255 255 255 255 255]
[255 255 255 255 255]
[255 255 255 255 255]]

Solution

1. Step 1: Understand erosion on the given image

   Erosion removes pixels at edges of white regions. The center pixel (0) surrounded by 255s will cause erosion to shrink the white area.

2. Step 2: Apply 3x3 kernel erosion

   Since the kernel covers neighbors, any pixel with a zero neighbor becomes zero. The white cross shape will erode to a smaller cross with zeros at the center and edges.

3. Final Answer:

   White cross with zeros at center and edges as shown in option A -> Option A

4. Quick Check:

   Erosion shrinks white, so edge pixels vanish but some inner pixels remain [OK]

Hint: Erosion removes edge pixels, shrinking white areas [OK]

Common Mistakes:

• Assuming erosion keeps center pixels
• Confusing erosion with dilation output
• Ignoring zero pixels in kernel neighborhood

4. You wrote this code to perform opening on an image img but it does not remove noise as expected:

kernel = np.ones((3,3), np.uint8)
opened = cv2.dilate(cv2.erode(img, kernel), kernel)

What is the error and how to fix it?

medium

A. Kernel size is too small; increase kernel size to remove noise.

B. The order is reversed; opening is dilation followed by erosion, so swap the calls.

C. Use cv2.morphologyEx with cv2.MORPH_OPEN instead for correct opening.

D. The order is reversed; opening is erosion followed by dilation, so code is correct.

Solution

1. Step 1: Check the definition of opening

   Opening is erosion followed by dilation. The code applies erosion then dilation, which is correct in order.

2. Step 2: Identify practical issue

   Manual calls may work but can be error-prone; using cv2.morphologyEx with MORPH_OPEN is recommended for correct and optimized opening.

3. Final Answer:

   Use cv2.morphologyEx with cv2.MORPH_OPEN instead for correct opening. -> Option C

4. Quick Check:

   Use built-in morphologyEx for opening [OK]

Hint: Use cv2.morphologyEx with MORPH_OPEN for opening [OK]

Common Mistakes:

• Swapping erosion and dilation order
• Not using built-in morphology functions
• Assuming kernel size fixes logic errors

5. You have a noisy binary image with small black holes inside white objects. Which morphological operation should you apply to fill these holes without changing the object shapes much?

hard

A. Closing

B. Dilation

C. Opening

D. Erosion

Solution

1. Step 1: Understand the problem of black holes inside white objects

   Black holes are small dark spots inside white regions that need to be filled.

2. Step 2: Choose operation that fills holes without shrinking objects

   Closing is dilation followed by erosion; it fills small holes and gaps while preserving object shape.

3. Final Answer:

   Closing -> Option A

4. Quick Check:

   Closing fills holes inside white objects [OK]

Hint: Closing fills holes inside white objects [OK]

Common Mistakes:

• Using erosion which shrinks objects
• Using opening which removes noise but not holes
• Confusing dilation alone with closing